Re: [PATCH 1/2] arch/ia64/kvm/kvm-ia64.c: Add local_irq_restore inerror handling code

From: Julia Lawall
Date: Tue Jul 22 2008 - 15:18:43 EST


On Tue, 22 Jul 2008, Simon Holm Thøgersen wrote:

> tir, 22 07 2008 kl. 18:44 +0200, skrev Julia Lawall:
> > From: Julia Lawall <julia@xxxxxxx>
> >
> > There is a call to local_irq_restore in the normal exit case, so it would
> > seem that there should be one on an error return as well.
> >
> > The semantic patch that makes this change is as follows:
> > (http://www.emn.fr/x-info/coccinelle/)
> >
> > // <smpl>
> > @@
> > expression l;
> > expression E,E1,E2;
> > @@
> >
> > local_irq_save(l);
> > ... when != local_irq_restore(l)
> > when != spin_unlock_irqrestore(E,l)
> > when any
> > when strict
> > (
> > if (...) { ... when != local_irq_restore(l)
> > when != spin_unlock_irqrestore(E1,l)
> > + local_irq_restore(l);
> > return ...;
> > }
> > |
> > if (...)
> > + {local_irq_restore(l);
> > return ...;
> > + }
> > |
> > spin_unlock_irqrestore(E2,l);
> > |
> > local_irq_restore(l);
> > )
> > // </smpl>
> >
> > Signed-off-by: Julia Lawall <julia@xxxxxxx>
> >
> > ---
> > arch/ia64/kvm/kvm-ia64.c | 9 +++++++--
> > 1 files changed, 7 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/ia64/kvm/kvm-ia64.c b/arch/ia64/kvm/kvm-ia64.c
> > index 318b811..14e5c99 100644
> > --- a/arch/ia64/kvm/kvm-ia64.c
> > +++ b/arch/ia64/kvm/kvm-ia64.c
> > @@ -125,8 +125,10 @@ void kvm_arch_hardware_enable(void *garbage)
> > PAGE_KERNEL));
> > local_irq_save(saved_psr);
> > slot = ia64_itr_entry(0x3, KVM_VMM_BASE, pte, KVM_VMM_SHIFT);
> > - if (slot < 0)
> > + if (slot < 0) {
> > + local_irq_restore(saved_psr);
> > return;
> > + }
> > local_irq_restore(saved_psr);
>
> If this change makes sense, how about merging the two local_irq_restore
> calls and put them before the if? And shouldn't your checker be able to
> figure this?

No, it pretty much only does what it's told, which is to put a
local_irq_restore before the return.

Since slot is a local variable, it does seem like the call to
local_irq_restore could be safely moved upwards, so I will do that
instead.

thanks,
julia